Can I bring my existing Spectral ruleset to Karate?

Ruleset conversion and the ingest of existing Spectral or vacuum findings as evidence are shipping this quarter. The graded evidence join, where rules read your coverage graph and contract-test results rather than the document, is available today.

What can Karate check that Spectral cannot?

Three questions that are not in the document at all: whether an operation is claimed by a requirement, whether any run has ever exercised it, and whether anything other than your own mock has ever answered it. A Spectral ruleset cannot ask these, not because of a gap in the tool but because the answers live in execution data rather than in the file.

Is the Karate grade deterministic like a Spectral lint?

Yes. There is no model and no network call in the grade. The same document and the same rulepack produce the same findings, the same dimension scores, the same ship verdict and the same maturity level, every time. That is what makes it attestable rather than something to take on trust.

Karate vs Spectral

Q: Is Karate a replacement for Spectral?

No, and it is not designed to be. Spectral is a good document linter and it is free. Karate is the layer above it: it runs its own deterministic rulepack, and then grades the execution evidence a document linter cannot see. The intended path is to keep the linting you already have and add the evidence join on top.

Spectral is a good document linter and it is free. It is also the ceiling: a ruleset can only ever ask questions the document can answer about itself.

Karate runs its own deterministic rulepack and then grades the execution evidence behind the document, into a ship verdict and a maturity level.

The Short Version

Bring your ruleset, keep your findings, add the evidence

Bring your ruleset

Conversion onto an engine whose rules can also read your coverage graph. Shipping this quarter.

Keep your findings

Existing Spectral or vacuum output ingested as evidence rather than discarded. Shipping this quarter.

Add the evidence

Rules that read your coverage graph and contract-test results, graded into a verdict. Available today.

Side by Side

Where the two overlap, and where they don't

Capability	Karate	Spectral
Lints an OpenAPI document	Yes	Yes
Custom rulesets	Yes	Yes, and widely adopted
Severity, JSON pointer and line on findings	Yes	Yes
Deterministic, no model in the result	Yes	Yes
Was this operation ever called by a run	Yes, from the coverage graph	Not answerable from a document
Is this operation claimed by a requirement	Yes	Not answerable from a document
Did anything but a mock ever answer it	Yes	Not answerable from a document
Scored dimensions and a ship verdict	Five dimensions, three verdicts	Findings list
Named API maturity level	Six levels	Not in scope

The rows where Spectral reads "not answerable from a document" are not shortcomings of the ruleset. The answers are simply not in the file, at any quality of rule.

Choosing

When each one is the right answer

Spectral is the right answer when

✓You want a free, well-supported document linter in CI
✓Your goal is enforcing a house style across many specs
✓You have no execution evidence to join to yet

Add Karate when

✓A green lint is no longer convincing anyone
✓You need to know which operations nobody calls
✓Someone has to sign off that the API is ready to ship
✓You want the same tool to run the tests that produce the evidence

The honest summary: a document linter establishes that your API is well described. Nothing about a well-described API tells you it is exercised, and that is the gap the grade exists to close.